At a glance. The AI Strategy for the Federal Public Service 2025-2027 sets direction for the adoption of artificial intelligence (AI) across the Government of Canada. The first public AI Register, published in November 2025, lists more than 400 systems across 42 institutions, from early research to deployed tools. The Strategy sets the destination; delivery discipline decides how many of those systems become dependable public services.
What the Strategy sets out
The Treasury Board of Canada Secretariat (TBS) launched the AI Strategy for the Federal Public Service on March 4, 2025. It is the first AI strategy for the federal public service. It followed engagement from May to October 2024, and the government aims to update it every two years [1].
The Strategy rests on four principles: human-centred, collaborative, ready and responsible [2]. “Ready” is the principle delivery teams should read most closely. It calls for infrastructure that is durable and protected, reliable data and skilled staff [2]. Those are the conditions that decide whether a promising pilot can run as a service.
Action is organized under four priority areas [1]:
| Priority area | Focus | Implication for delivery teams |
|---|---|---|
| AI Centre of Expertise | Support and coordinate government-wide AI work | Use shared capacity before building alone |
| Security and responsible use | Keep systems secure and used responsibly | Put compliance work in the schedule |
| Training and talent | Skills and development pathways | Budget for people, not only platforms |
| Openness and transparency | Public trust in how AI is used | Publish what systems do and achieve |
Where federal AI stands today
On November 28, 2025, TBS published the first public register of AI uses in the federal government. It includes input from 42 institutions and more than 400 systems, ranging from early research and proof-of-concept projects to fully deployed tools [3]. Each entry describes the system’s purpose, its intended or current use, and whether it was built in-house or by a vendor. AI embedded in low-risk commercial products, such as spell checkers, is excluded [3]. TBS presents the register as part of the Strategy and plans public consultations on it in 2026 [3].
For delivery leaders, the register is a portfolio view. The useful question is not how many systems exist. It is how many will move from exploration to dependable service, and on what evidence.
Where the Strategy meets the Directive
The Strategy sits on top of existing rules. The one that matters most for delivery is the Directive on Automated Decision-Making. It applies to automated decision systems in production and excludes systems operating in test environments [4].
That makes the move to production a compliance event. Before a system that makes or supports administrative decisions goes live, the department must complete the Algorithmic Impact Assessment (AIA) and release its results [4]. Depending on the impact level, it may also need peer review, staff training and approval by the deputy head or by Treasury Board [4].
None of this is difficult when planned. All of it is difficult when discovered in the final month of a project. My own work is the delivery of Industry 4.0 programs, spanning industrial Internet of Things (IIoT), manufacturing execution systems (MES) and manufacturing intelligence. The setting differs from government, but the pattern is the same: a pilot proves that something can work, and it does not prove that an organization can run it.
Five delivery disciplines
The following framework turns the Strategy’s direction into delivery practice. Each discipline maps to at least one priority area.
- Define the outcome before the pilot. State the service problem, the current baseline and the measure of success. A processing-time or error-rate baseline taken before the pilot is worth more than any demonstration. (Openness and transparency.)
- Set stage gates from pilot to production. Agree in advance what evidence moves a system forward: performance against the baseline, a completed AIA, the requirements for its impact level, a named operational owner and a funded run budget. (Security and responsible use.)
- Use shared capacity first. The AI Centre of Expertise exists to support and coordinate government-wide AI work [1]. Teams that reuse common tools and lessons spend their budget on the service problem rather than the plumbing. (AI Centre of Expertise.)
- Plan the workforce side. Training and talent is one of the four priority areas [1], and the Directive scales training requirements with impact level [4]. Treat staff readiness as a deliverable with an owner and a date. (Training and talent.)
- Report outcomes, not activity. The register records what exists [3]. Departments should also be able to show what each system changed for clients, at what cost and with what error rate. (Openness and transparency.)
A useful test for any AI initiative is whether its project plan contains a task, an owner and a date for each of these five. If one is missing, that is where the schedule risk sits.
How other governments connect strategy to delivery
In the United States, Office of Management and Budget (OMB) Memorandum M-25-21 required each agency to designate a Chief AI Officer within 60 days. It required agencies covered by the Chief Financial Officers Act to develop AI strategies within 180 days and make them public, and it requires agencies to publish AI use case inventories at least annually [5].
In the United Kingdom (UK), the AI Playbook published by the Government Digital Service sets ten principles, including managing the full AI life cycle and working with commercial colleagues from the start [6]. The UK’s Algorithmic Transparency Recording Standard had 59 published records by May 2025 [7].
At the multilateral level, the AI Principles of the Organisation for Economic Co-operation and Development (OECD) include a recommendation to build human capacity and prepare for labour market transition [8].
| Jurisdiction | Central instrument | Delivery mechanism |
|---|---|---|
| Canada | AI Strategy 2025-2027 | AI Centre of Expertise; public AI Register |
| United States | OMB Memorandum M-25-21 (2025) | Chief AI Officers; agency AI strategies; annual inventories |
| United Kingdom | AI Playbook (2025) | Ten principles; mandatory transparency records |
| OECD | AI Principles (2019, updated 2024) | Shared recommendations to governments |
The instruments differ in form. The direction is shared: a central point of expertise or accountability, a public account of what AI is in use, and explicit attention to skills.
The bottom line
The Strategy and the register give the federal public service a direction and a baseline count. What turns them into results is ordinary delivery discipline: outcomes defined up front, gates that require evidence, shared capacity, a prepared workforce and honest reporting. None of it is new. All of it is easier to build in at the start than to retrofit at launch.
References
- Treasury Board of Canada Secretariat, “Canada launches first-ever Artificial Intelligence Strategy for the federal public service,” news release, Waterloo, ON, Canada, Mar. 4, 2025.
- Treasury Board of Canada Secretariat, “AI Strategy for the Federal Public Service 2025-2027,” Government of Canada, Mar. 2025.
- Treasury Board of Canada Secretariat, “Canada launches first register of AI uses in federal government,” news release, Ottawa, ON, Canada, Nov. 28, 2025.
- Treasury Board of Canada Secretariat, “Directive on Automated Decision-Making,” Government of Canada, 2019, as amended.
- Office of Management and Budget, “Accelerating Federal Use of AI through Innovation, Governance, and Public Trust,” Memorandum M-25-21, Executive Office of the President, Washington, DC, USA, Apr. 3, 2025.
- Government Digital Service, “Artificial Intelligence Playbook for the UK Government,” GOV.UK, Feb. 10, 2025.
- E. Hess-Rheingans, A. Hart, S. Handyside and S. Beaugrand, “Making the Algorithmic Transparency Recording Standard (ATRS) mandatory across government,” Data in Government blog, Government Digital Service, May 8, 2025.
- OECD, “OECD AI Principles” (Recommendation of the Council on Artificial Intelligence, OECD/LEGAL/0449), adopted May 2019, updated May 2024.